Beruflich Dokumente
Kultur Dokumente
I.
INTRODUCTION
A straightforward industrial development process of electrical variable speed drive systems is fortunately assisted by modern commercial calculation tools [1-3]. However, the extensive and almost precise analysis of the interaction of different control algorithms, like the vector control or the brushless DC mode, within introduced innovative permanent magnet motor topologies, as e.g. the higher harmonic air-gap based design, is still a big challenge [4-6]. Several simplifications have to be done in order to treat the analysis in a reasonable way. The implemented main software features, the interacting converter hardware and the motor itself are almost analyzed in the time-domain.
dc link T1
dc to ac power converter T2 T3
energy conversion
C pm-motor T4 T5 T6 load
voltage measuring
current measuring
sensor
interface communication
interface
interface
FEPROM
Fig.1. Outline of important devices of the considered drive system. Manuscript received June 23, 2008. Christian Grabner is with the Research and Development Department of Drive Systems, Siemens AG, Frauenauracherstrasse 80, D-91056 Erlangen, Germany, (phone: +49 162 2515841, e-mail: grabner.christian@siemens.com).
Fig.3. Permanent magnet assisted rotor (left) and novel stator topology with a three phase winding system (right).
The applied principle of synchronous energy conversion is based on the steady interaction of two separate parts stator and rotor of same rotational speed and a coinciding even number of alternating electromagnetic poles along the circumferential air gap between both considered parts [10]. The performed motor research is based on alternative possibilities for the generation of desired stator flux density waves within the air gap. The moving rotor with attached permanent magnets is characterized by its even number of involved magnetic poles, whereas the fixed novel stator winding design comprises the same pole number by means of higher space harmonics. The suggested un-skewed stator and rotor topology is thereby suitable to induce continuously electromechanical torque with very low ripple. Do to the axially un-skewed lamination, only the 2D triangular finite element representation as shown in Fig.4 is necessary [1]. The rotor moves on stepwise during the transient calculation with the aid of the applied band technique [2,3]. Thus, elements which are situated at these band domain may be deformed during motion since some of their nodes may follow the moving rotor reference frame and some the stationary stator part. These element distortions are occurring in dependency on the rotational movement. Thus, the continuous element distortion enforces a re-meshing of air-gap elements lying on the band. Governing 3D stator end winding effects are commonly included in the 2D finite element calculation by external parameters within directly coupled electrical circuits, as shown in Fig.5. The applied voltage can be arbitrarily depending on time. This suggests coupling the output terminals of the power converter topology from Fig.1 directly with each stator phase to complete the power flow path. The actual rotor position is always accessible within the finite element calculation, so that basic control features could be implemented advantageously. The controlled vector or brushless DC mode of the permanent magnet motor is preferable treated with numerical methods in the time-domain [3]. As a consequence, eventually occurring higher harmonics in the calculated time courses of current or torque which are succeeding limiting values could be monitored in advance. Thus, further costs for extensive prototyping could be avoided.
Fig.4. Finite element mesh of rotor and stator laminations, magnets, and stator winding system within the plane xycoordinate system.
Fig.5. Coupling of the 2D finite-element-model to one stator phase circuit with external parameters which are representing 3D end winding effects.
0,63
1,26
1,88
2,51
3,14
3,77
4,40
5,03
5,65
6,28
0,63
1,26
1,88
2,51
3,14
3,77
4,40
5,03
5,65
6,28
Fig.7. Calculated radial flux density component along the circumferential air gap distance at no-load.
1,0 0,9 radial airgap flux density [T]
Fig.10. Calculated radial flux density component along the circumferential air gap distance at default stator current.
1,0 0,9 radial airgap flux density [T] 0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0
0,8 0,7 0,6 0,5 0,4 0,3 0,2 0,1 0,0 0 7 14 21 28 35 42 49 56 63 70 harmonic order [1]
14
21
28
35
42
49
56
63
70
Fig.11. Fourier spectrum Bk of the radial flux density component at default stator current.
0,003
0,006
0,009
0,012
0,017
0,020
0,023
0,026
0,029
Fig.12. Calculated time-dependent phase-to-phase voltage for a constant speed of 150 rpm.
Fig.13. Measured time-dependent phase-to-phase voltage for a constant speed of 150 rpm. One division corresponds to 10 ms in the abscissa and 2 V in the ordinate.
C. Induced no-load voltage at constant speed Irregularities in the air gap caused by slots and magnetic inhomogenities in the magnetic material tend to produce higher harmonics in the induced voltage in simple winding designs [10]. The proposed un-skewed motor construction of 140 mm axial length with surface magnets and a single tooth winding system overcomes such circumstances and delivers a sinusoidal stator voltage shape as shown in Fig.12. With the number of p = 14 poles and a constant speed of 150 rpm, the fundamental supply frequency is f = np = 35 Hz and the relevant fundamental period is 28.6 ms. A comparison of the numerical calculated results with the measured voltage in Fig.13 shows a very good correlation. D. Cogging torque effects at constant speed Since the demands for torque uniformity is essential, these detrimental factors must be taken into account during the machine design in order to reduce undesired cogging torque effects. Thus, negative impacts on the accuracy of the shafts even speed or excessive acoustic noise emission could be prevented. This becomes more crucial, when unskewed stator and rotor parts are used [10]. The calculated cogging torque at generator no-load operation in case of the un-skewed motor design in Fig.3 is exemplarily shown in Fig.14.
k of the wide spectrum in Fig.11 are obtained. There exists a lot of ordinal numbers with even different magnitudes. However, only the invoked fourteenth component with B14 = 0.045 T can
interact with the rotor part in order to generate a constant mechanical average torque. Other harmonic waves are therefore undesired and cause losses. In order to restrict such effects, the geometric air gap distance is limited to be minimal of about 0.34 mm. Due to the used surface magnet design the magnetic air gap distance is wide enough to restrict additional losses within the rotor magnets due to excessively eddy current effects. This is also obviously from Fig.9 due to the very weak penetration of the stator field into the magnets and the rotor yoke.
IV. VECTOR CONTROLLED HIGHER HARMONIC AIRGAP-WAVE BASED PERMANENT MAGNET MOTOR The vector control mode operates the permanent magnet motor like a current source inverter driven machine applying a continuous current modulation [11-13]. Therefore, a very precise knowledge of the rotor position is necessary. This could be managed by applying an encoder on the rotating shaft. The vector control software is adapted to the hardware system which is schematically depicted in Fig.2. A. Mathematical motor model The practical realization of the control structure is fortunately done within the rotor fixed (d, q ) reference system, because the electrical stator quantities can there be seen to be constant within the steady operational state [14,15]. The stator voltage and flux linkage space vectors are therefore formulated in the (d, q ) rotor reference frame as d Sdq ( ) uSdq ( ) = rS iSdq ( ) + + j ( ) Sdq , (2) d Sdq ( ) = lS iSdq ( ) + Mdq , (3) whereby rS denotes the normalized stator resistance and () stands for the mechanical speed of the shaft. Thereby, the relation (3) covers only weaksaturable isotropic motor designs. The inclusion of some basic identities for the permanent magnet flux space vector in the (d, q ) system, namely d Mdq = 0 , Mdq = M + j 0 , (4) d reduces the system (2), (3) to the set of equations
Fig.15. Measured cogging torque for a constant speed of 10 rpm. One division corresponds to 70 ms in the abscissa and 0.05 Nm in the ordinate.
0,16 0,14 0,12 cogging torque [Nm] 0,10 0,08 0,06 0,04 0,02 0,00 0 7 14 21 28 35 42 frequency [Hz] 49 56 63 70
(5) (6)
Unfortunately, both equations (5), (6) are always directly coupled without the exception of standstill at = 0 . That fact is very unsuitable in particular for the design of the current controller. Thus, with regard to the used vector control topology, a more favorable rewritten form of (5), (6) as
ud ( ) = uSd ( ) + lS iSq ( ) =
(7)
(8)
is commonly introduced.
ndef
ndem
id,dem=0
nact
I2T monitoring
id,act
iq,act
PT1 smoothing
id,act
d,q a, b, c D conversion
imeas
PT1 smoothing
nmeas
Fig.17. Outline of important devices of the vector controlled permanent magnet motor.
i (t ) = I k sin (2kf t + k )
k =1
(9)
of the calculated electrical current from Fig.19 leads I the harmonic components k , k . It is obvious from the spectrum in Fig.21, that only the desired fundamental component 1 = 3.75A at 93.3 Hz is I governing the total spectrum. The vector control in conjunction with the special motor obviously avoids additional harmonic components and restricts therefore unexpected thermal heating due to timeharmonic currents. D. Mechanical torque and harmonic spectrum The numerical calculated time-dependent mechanical torque m(t) is depicted in Fig.22. It results from the series expansion in time
m (t ) =
M
k =0
sin (2kf t + k )
(10)
that there exists within the harmonic components Mk , k 0 almost the desired constant contribut ion of M = 1.3Nm . By taking a closer look to the
0
calculated mechanical torque in Fig.23, a very dist inct undesired harmonic component M6 = 0.1Nm known as load pulsation moment exists even in spite of the novel un-skewed motor topology at the frequency of 559.8 Hz.
10,0 7,5
5,0 2,5 0,0 -2,5 -5,0 -7,5 -10,0 0,066 0,068 0,070 0,072 0,074 0,076 time [s] 0,079 0,081 0,083 0,085 0,087
ud,dem
usd,dem
ls isq,act
Fig.18. Block diagram of the decoupling-circuit.
Fig.19. Calculated motor current i(t) for a speed range of 400 rpm and a constant load of 1.35 Nm.
Fig.20. Measured motor current i(t) for a speed range of 400 rpm and a constant load of 1.35 Nm. One division corresponds to 2 ms in the abscissa and 2.5 A in the ordinate.
4,0 3,5 3,0 load current [A] 2,5 2,0 1,5 1,0 0,5 0,0
Fig.21. Fourier coefficients k of the motor current for a speed I range of 400 rpm and a constant load of 1.35 Nm.
1,6 1,4 1,2 load torque [Nm] 1,0 0,8 0,6 0,4 0,2 0,0 0,066
The machine equations in terms of according stator space vector notation is written in a stator fixed reference frame as d S ( ) uS ( ) = rS iS ( ) + . (11) d The introduced stator flux space vector in (11) is fortunately written in case of weak-saturable isotropic inductances as S ( ) = lS iS ( ) + M , (12) whereby a rotor flux space vector of constant magnitude is assumed. The governing relation for the brushless DC feature is derived from (11),(12) as d uS ( ) = rS iS ( ) + lS iS ( ) + j M , (13) d whereby any transformation into a rotor fixed (d, q ) reference system is avoided. B. Simplified block diagram of the closed-loop brushless DC control The control in Fig.24 is realized with an outer speed and an innermost current control cascade. The imported Hall sensor signals, denoted after the interface with meas, are transformed to a continuous actual speed value with the aid of a D-element. Due to the low number of six Hall sensors, only very rough speed detection is feasible. The smoothed signal by a PT1 is denoted as nact. The actual speed nact is compared with the demanded speed ndem. The difference is in Fig.24 applied to the moderate PI speed controller which delivers the required motor current magnitude. This value is further limited by the thermal I2T protection, which takes implicit use of the actual measured motor current iact.
0,068
0,070
0,072
0,074
0,079
0,081
0,083
0,085
0,087
Fig.22. Calculated electromagnetic torque m(t) for a speed range of 400 rpm and a constant load of 1.35 Nm.
1,6 1,4 1,2 load torque [Nm] 1,0 0,8 0,6 0,4 0,2 0,0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 frequency [Hz]
Fig.23. Fourier coefficients Mk of the shaft torque for a speed range of 400 rpm and a constant load of 1.35 Nm.
PI controller
idem nact
PI controller
udem
I2T monitoring
PT1 smoothing
nmeas
D conversion
meas
Fig.24. Simplified block diagram of the closed-loop speed and current control.
The innermost loop in Fig.24 serves as current control. The actually measured motor current imeas first passes a PT1 smoothing block, before the obtained signal iact is processed with an 8 kHz sampling rate of the current-controller. Depending on the difference between the demanded current magnitudes idem and the smoothed measured current iact, the necessary motor voltage magnitude udem is calculated with the aid of a PI current controller. C. Time dependent current shape and harmonic spectrum The influence of the electrical current commutation from one phase to the other can be clearly seen in the calculated time-dependent course of Fig.25 for rated-load and the speed of 400 rpm. The measured quantity is given in Fig.26. Direct comparison of calculated with measured courses shows a very good concordance. The implemented numerical analysis is also very suitable to predict even higher harmonics in the motor current. The complete current spectrum (9) in Fig.27 contains the fundamental component 1 = 3.84A at 93.3 Hz. Moreover, I some very distinct peaks within the spectrum could be observed. Several invoked higher harmonics such as 5 = 0.81A at 466.6 Hz, 7 = 0.62A at 653.3 Hz, I I = 0.31A at 1026.3 Hz and = 0.21A at I I
11 13
known undesired torque fluctuations within Fig.28 can be seen more clearly in the harmonic spectrum depicted in Fig.29. Both distinct higher components are found to be M6 = 0.12Nm at 559.8 Hz and M = 0.09Nm at 1119.6 Hz. Other contributions to
12
the torque ripple are obviously suppressed. The occurring undesired load tip effects are well known to be responsible for eventually undesired noise emission.
6,0 4,5 3,0 load current [A] 1,5 0,0 -1,5 -3,0 -4,5 -6,0 0,057
0,059
0,061
0,063
0,065
0,070
0,072
0,074
0,076
0,078
Fig.25. Calculated motor current i(t) for a speed range of 400 rpm and a constant load of 1.35 Nm.
1212.9 Hz are causing negative effects, such as undesired thermal heating. D. Time dependent torque and harmonic spectrum The numerical calculated time-dependent mechanical torque m(t) is depicted in Fig.28 for ratedload and a speed value of 400 rpm. It results from the series expansion (10) that there exists within the harmonic components Mk , k 0 almost the desired constant contribution of M = 1.36Nm . The
0
Fig.26. Measured motor current i(t) for a speed range of 400 rpm and a constant load of 1.35 Nm. One division corresponds to 2 ms in the abscissa and 1.5 A in the ordinate.
losses are still reducing the thermal torque speed characteristic. In order to overcome those circumstances within certain ranges, the complex vector mode is commonly preferred. VII. CONCLUSION The novel axially unskewed higher harmonic airgap wave based permanent magnet motor technology has been analyzed with regard to the closed-loop vector as well as brushless DC control method. The main focus is thereby given to the verification of previously unknown and almost undesired effects, which could significantly worsen the quality of the complete drive system. The vector control method enforces only one distinct fundamental component in the electrical current consumption, whereas the brushless DC control causes a wide harmonic current spectrum. Undesired load pulsation effects of the unskewed motor could be slightly reduced by preferring the vector control method. The applied transient electromagnetic-mechanical finite element calculation method with additionally coupled external circuits in the time-domain allows the inclusion of basic control features and is therefore very suitable for a straightforward and accurate analysis of the complete converter fed speed-variable drive system in advance. REFERENCES
[1] J.S. Salon, Finite element analysis of electrical machines, Cambridge University Press: Cambridge, 1996. [2] M.J. DeBortoli, Extensions to the finite element method for the electromechanical analysis of electrical machines, PhD Thesis, Rensselaer Polytechnic Institute, New York, 1992. [3] J.P.A. Bastos and N. Sadowski, Electromagnetic modeling by finite element method, Marcel Dekker: New York/Basel, 2003. [4] P.F. Brosch, Moderne Stromrichterantriebe, Vogel: Wrzburg, 1989. [5] W.Leonhard, Control of electrical drives, Springer: Berlin, 2001. [6] G.K. Dubey, Fundamentals of electrical drives, Alpha Science Int.: Pangbourne, 2001. [7] K.B. Bimal, Power electronics and variable frequency drives: technology and applications, IEEE Press: New York, 1996. [8] K. Heumann, Principles of power electronics. Berlin: Springer Verlag, 1986. [9] G. Seguier and G. Labrique, Power electronic converters: dc to ac conversion. Berlin: Springer Verlag, 1989. [10] T. Bdefeld and H. Sequenz, Elektrische Maschinen, Springer: Wien/New York, 1942. [11] P. Vas, Vector control of AC machines, Oxford University Press: Oxford, 1990. [12] K.G. Bush, Regelbare Elektroantriebe: Antriebsmethoden, Betriebssicherheit, Instandhaltung, Verlag Pflaum: Mnchen, 1998. [13] W. Nowotny and T.A. Lipo, Vector control and dynamics of AC drives, Clarendon Press: Oxford, 2000. [14] P. Vas, Electrical machines and drives: A space-vector theory approach, Clarendon Press: Oxford, 1996. [15] P. Vas, Parameter estimation, condition monitoring, and diagnosis of electrical machines, Clarendon Press: Oxford, 1993. [16] R. Lehmann, Technik der brstenlosen Servoantriebe, Elektronik, Vol. 21, 1989. [17] T.J.E. Miller, Brushless permanent magnet and reluctance motor drives, Clarendon Press: Clarendon, 1989. [18] J.R. Hendershot and T.J.E. Miller, Design of brushless permanent-magnet motors, Oxford university press: Oxford, 1994.
Fig.27. Fourier coefficients k of the motor current for a speed I range of 400 rpm and a constant load of 1.35 Nm.
1,6 1,4 1,2 load torque [Nm] 1,0 0,8 0,6 0,4 0,2 0,0 0,057
0,059
0,061
0,063
0,065
0,070
0,072
0,074
0,076
0,078
Fig.28. Calculated electromagnetic torque m(t) for a speed range of 400 rpm and a constant load of 1.35 Nm.
1,6 1,4 1,2 load torque [Nm] 1,0 0,8 0,6 0,4 0,2 0,0 0 200 400 600 800 1000 1200 1400 1600 1800 2000 frequency [Hz]
Fig.29. Fourier coefficients Mk of the shaft torque for a speed range of 400 rpm and a constant load of 1.35 Nm.
VI.
In case of the well-balanced motor construction and a diligent sensor adjustment, the mostly undesired torque fluctuation of the unskewed motor design still exist within the vector and brushless DC operational mode. In case of some technical applications, this disadvantage could be accepted; otherwise the rotor has to be skewed. However, for many circumstances, the easy unskewed motor construction and the much cheaper brushless DC control mode in conjunction with the higher harmonic air-gap wave based motor design is often favorized. Unfortunately, additional copper losses and varying iron